#16: ML from the Browser; Classifying the Social Web

1. Speaker: Amir Tabakovic

Abstract: This talk shows BigML makes machine learning more accessible via a well-defined workflow, insightful visualizations, and fully featured REST API. Using only a browser, we will develop a system to predict low risk loans using the rich data available from Lending Club. Techniques applied will include dataset transformations, random decision forests, clustering, anomaly detection, batch predictions, evaluations and more. Some background information will be provided how BigML is preforming anomaly detection.

Bio: 12 years in digital financial services and technology working currently for BigML. Result-oriented former internet start-up entrepreneur. Leading a market development team of business developers and mobile commerce sales people. MS in Business Administration from University of Bern.

2. Speaker: Sharon Hüffner

Abstract: This talk is about cloud-based brand protection and compliance for enterprise social media accounts. That means sorting through a lot of data: posts both from the brand itself and its commenters on multiple social networks like Twitter, Facebook, and LinkedIn. We use ML to build classifiers that detect everything from customer complaints to malicious URLs. But to do that, the first step is to discover and classify all the accounts associated with a particular brand. In this talk I'll give a quick overview of the (mostly text-based) machine learning we do at Nexgate, and focus on the problem of distinguishing corporate accounts from personal accounts. As is often the case, our emphasis is on data collection and feature engineering, with standard, state-of-the-art machine learning methods.

Bio: PhD in Mathematics from the Freie U. in Berlin, researching modularity in networks. Before that, BSc and MSc from Tel Aviv University, Israel. I've been living in Berlin for six years now, working for Nexgate (now Proofpoint) for four, now in the role of a Senior Data Scientist.

----

Thanks to NVIDIA for co-sponsoring the event!

http://photos3.meetupstatic.com/photos/event/6/f/3/600_442921779.jpeg